About Software Architecture and Planning
An agent that starts coding from a one-line request will make design choices on its own, and those choices are costly to undo. The skills here move the thinking earlier: interview you about the goal, write a spec, record the trade-offs and break the work into ordered tasks.
The category has two halves. Architecture skills cover system design, clean architecture, domain-driven design, API and GraphQL design, microservices and data-intensive systems, many of them distilled from well-known books. Planning skills cover interviews, specs, implementation plans, ticket breakdowns and architecture decision records, so the reasoning stays in the repository after the session ends.
When choosing, match the skill to the size of the decision. A quick API design checklist is enough for a single endpoint, while a new service calls for a skill that produces diagrams, written trade-offs and a decision record. Look at what the skill outputs and where it saves it, since a spec in a tracker and a spec in a markdown file feed different workflows. Book-based rule sets work best as background guidance and may need trimming to fit your codebase.
A good starting pair is one interview or spec skill and one architecture skill that fits your stack, with a task breakdown skill to hand the result to an implementer.